1. Field of the Invention
The present invention relates generally to methods and apparatus for improving the performance of file management within nonvolatile memory devices and particularly to increasing the speed of writing or storing information to such nonvolatile memory devices.
2. Description of the Prior Art
Various prior art methods and techniques were used to manage files, i.e. store data and read back data, within nonvolatile memory devices. Generally, a host device commands a controller, coupled between the host device and nonvolatile memory, to store certain information within nonvolatile memory and later to read the same. Such information depends on the application of the nonvolatile memory device. For example, in the case of digital cameras, digital pictures or photos is the information stored and retrieve from nonvolatile memory. In the case of Personal Computers (PCs), information is stored and retrieved from hard disk.
Since information is typically organized in sectors, each sector including a predetermined number of user data and a predetermined number of overhead data, the host commands the controller to store sector information by referencing addresses associated with particular sectors. For example, there may be sectors 0 through N and each a group of sectors may define a block which would also have an address associated therewith for identifying the same. The controller uses sector addresses to organize digital information within the nonvolatile memory device.
In one prior art technique, as a part of file management of nonvolatile memory, when the host device commands the controller to write or store information to one or more particular sectors, identified by logical block addresses (LBAs), the controller writes to physical block addresses (PBAs) in the nonvolatile memory. Each block includes a very large amount of nonvolatile memory space, for example, 64 Kbytes. When a particular sector is updated or re-written thereto in nonvolatile memory, the controller writes the updated sector information to another location within the 64 Kbyte block space. To keep track of the current sector information, flags and address information are utilized and are updated by the controller to reflect the status of the sector. U.S. Pat. No. 5,341,330, issued on Aug. 23, 1994 to Wells et al. and entitled “Method For Writing to a Flash Memory Array During Erase Suspend Intervals” is an example of the teachings of such prior art technique. In the case where a particular sector is updated within a block, the sector location including previous information is marked ‘old’ utilizing a flag and the new or current sector location is marked ‘new’. Finally, when the block is full, i.e. no free or available location remains, a new block is used to store further updates to sectors and the old block is eventually erased prior to being re-utilized.
An example of the above discussion is perhaps better shown by reference to the example of FIG. 1 depicting a block 10 and a block 12, each of which include 64 Kbytes of storage area organized into sector locations for storing sector information. The number of sector locations included within a block is a function of the size of each sector. In the case, for example, where each sector includes 512 bytes, the number of sectors included within a block having 64 Kbytes is obviously 64×1024 divided by 512 or 128.
Referring still to FIG. 1, when the host writes to a sector location identified by LBA 0, the controller stores said information into 14 and associated flag(s) are set to ‘new’ the first time such a write or store operation takes place after erasure of the block 10. However, after following writes to the same sector, eventually, sector 0 at 14 will be set to ‘old’ indicating that the information stored therein is no longer current and that the controller should read another location to obtain the latest sector 0 information. This occurs when sector 0 is re-written or updated a following time and because information at 14 cannot be re-written without the block 10 first being erased. Since no erasure of block 10 has taken place, the next time sector 0 is written, its information will be placed at 28 and while the flag for location 14 will be set to ‘old’, the flag for location 28 is set to ‘new’ indicative of the most up-to-date sector 0 information until the latter is again updated, at which time the current information is stored for location 44 in block 10 and the flag at 44 is set to ‘new’ while the flag for location 28 is set to ‘old’.
The scenario described above applies to the writing or updating of all other sectors. By brief way of example, sector information identified by LBA 1, is initially written at 16 and the next time it is written, it is written to the next available location in Block 10 which is location 30 and the following time after that when it is written by the host, it is written at 36 and the flags of 16, 30 and 36 are updated as describe above. This process continues until the block 10 becomes full at which time a new, or available, or free block is found by the controller, in this case, block 12. From thereon, updated sector information is written to the block 12, not only this, but at some point, if necessary, all sector locations including current sector information are moved to the block 12, as explained in U.S. Pat. No. 5,341,330.
For example, in FIG. 1, after the first time when the sector identified by LBA 50 is written, assuming the host commands the controller to write to LBA 50 a next time and the block 10 is found to be full, the re-writing of sector 50 takes place within the block 12 rather than the block 10. In fact, the re-written sector 50 information is written at 50 and all other sectors designated as having current or ‘new’ sector information are moved to the block 12. This includes the sector identified by LBA 901, which is at 24 in block 10 and moved to 52 in block 12, the sector identified by LBA 902, which is at 26 in block 10 and moved to 54 in block 12, the sector identified by LBA 900, which is at 34 in block 10 and moved to 56 in block 12 (note that this sector was initially written at 22 but the sector information at 22 is now ‘old’ and the most recent information resides at 34, which is the reason for moving the information stored at 34 rather than the information at 22), the sector identified by LBA 1, which is at 36 in block 10 and moved to 58 in block 12 and so on.
The above prior art technique is described in further detail in U.S. patent application having Ser. No. 09/620,544 filed on Jul. 21, 2000 and entitled “Moving Sectors Within a Block of Information In a Flash Memory Mass Storage Architure”, the disclosure of which is incorporated herein by reference as though set forth in full. The problem with this technique is that to move all of the sectors including current information to another new block is time-consuming and therefore a performance hindrance. This problem is even further exaggerated when using smaller block sizes as there are more numerous move operations with smaller block sizes and smaller block sizes are more prevalent by today's users of nonvolatile memory devices, particularly by users of nonvolatile memory devices.
In the patent document referred to hereinabove, a method and apparatus is introduced for improving the performance of managing files or data within nonvolatile memory by organizing the memory into smaller block sizes and introducing a virtual logical block address (VLBA) to PBA relationship and a unique VLBA was assigned to each block and within each VLBA were sectors arranged in sequential order for decreasing the number of moves to expedite or improve the performance of the system through the use of mapping of PBAs to VLBAs. This VLBA to PBA mapping caused the size of the space manager within the controller device to decrease thereby resulting in a less expensive manufacturing of the controller device. However, in this method, it is presumed that sectors are written in sequential order by the host, if this is not the case, there is much wasted memory space.
In further explanation of prior art techniques, FIG. 1(a) shows another method for updating sector information in that when sector information is re-written by a host, the new or updated information need be written to a free block. For example, as shown in FIG. 1(a), when sector information, identified by LBA 0 in Block 0, is re-written or updated, the updated LBA 0 sector information is written to LBA 0 of Block 1. All other sectors within the Block 0 need then be moved to Block 1. Accordingly, every time there is a re-write or update of a sector, an entire block of information is moved to a new or free block. Obviously, this adversely affects system performance because every time there is a re-write of a sector, a new location within a free block is written thereto while the old information remains in the previous block until the system erases the latter.
In yet another prior art technique, sectors are not moved necessarily right away after every sector information update, rather, re-writes and move operations are kept track thereof and when a block is full or nearly full of mostly old sector information, its current sector information is then moved to a new block. For example, as shown in FIG. 1(b), when sector information to LBA 0 is updated, it is written to an available sector location in an available or free block but the remaining sectors of the previous block are not moved to the new block. Thus the previous block continues to hold some current sector information as well as some old sector information. The system keeps track of rewrites so that it has knowledge of which sectors are old and which are current and when a block is full or nearly full of old sector information, it moves the current sector information, if any, to the new of available block.
Thus, the need arises for a system and method for file or data management of information that is organized into sectors within nonvolatile memory devices while improving the performance for doing the same in an inexpensive manner.